Install flash-attn package and set default device to CUDA in app.py

app.py

@@ -29,6 +29,11 @@ import traceback
 # import torch
 from conversation import Conversation
 from transformers import AutoModel, AutoTokenizer, TextIteratorStreamer
+import subprocess
+
+subprocess.run('pip install flash-attn --no-build-isolation', env={'FLASH_ATTENTION_SKIP_CUDA_BUILD': "TRUE"}, shell=True)
+
+torch.set_default_device('cuda')
 
 logger = build_logger("gradio_web_server", "gradio_web_server.log")